Project Description

Medical Record Extraction

To extract ICD 10 & CPT codes and classify them based on text in comments.

Customer’s Objective

Extracting multiple forms of data gets challenging (ICD 10 Codes, CPT Codes etc). ICD-10 codes are alphanumeric codes used by doctors, health insurance companies, and public health agencies across the world to represent diagnoses. Every disease, disorder, injury, infection, and symptom have its own ICD-10 code. These codes are used for everything from processing health insurance claims to tracking disease epidemics and compiling worldwide mortality statistics.

Extracting codes from standard terminologies is a regular and indispensable task often encountered in medical and healthcare fields. Diagnosis codes, procedure codes etc. are all manually extracted from patient records by trained human coders.

Our Approach

We had developed a deep learning-based text extraction with an accuracy rate of 88% (and more), which had helped to detect boundaries and text associated with the boxes. Patterns matching and classification algorithm was used to classify extracted text for preconditions and trigger recommendations.

Outputs:

  • Coded Charts, Quality of Care, Suspect Conditions
  • Searchable and highlighted PDFs
  • Output in the form of XML, CSV, API

Business Impact

Operations savings in terms of human coders by way of accuracy and time.

RESULT

0% +
EXTRACTION ACCURACY